Wordle data mining based on deep learning

Abstract

In just a few months, Wordle has grown from a few players to several million users after "viral spread" by major social platforms. To study the secret of Wordle game success We construct a time-series based model by building a QL-LSTM model to explain the daily variation in the number of reported results and try to predict the game reports on March 1, 2023, which has a prediction interval of [21127, 24093], while we construct features for vowel and consonant digits, wordiness, number of letter repetitions, and word frequency. and assessed the relevance and significance of the percentage of those reporting being in hard mode. Only word frequency and percentage in hard mode had a slightly negative and significant relationship. The mystery of Wordle is unveiled by the above exploration.